Various techniques are known for extending the field of view (FOV) of a camera. One technique is to mount a plurality of cameras on circular or semi-circular mount structure to implement what is sometimes referred to as a camera ring. A camera ring allows for a panoramic view with much less “fish-eye” distortion and higher resolution than a single camera. With a camera ring, the individual cameras each produce images that overlap, and the overlapping images are then stitched together to produce a single, expanded FOV image.
Various techniques are also used to implement the image stitching. The most accurate technique is to generate depth fields and then apply the video images captured by the cameras as textures overlaying the depth fields. Although this technique is fairly accurate, the processing time associated with generating the depth fields results in excessive latency for real-time applications, such as driving a vehicle with indirect vision.
As may be appreciated, when implementing a camera ring, a technique is also used to transition between the cameras. One relatively simple technique is to simply switch between cameras at about the center of the angle between the two cameras. This has the advantage of low latency, but causes a rather pronounced “jump” in the scene due to parallax difference.
Hence, there is a need for a system and method for transitioning between cameras on a camera ring that eliminates, or at least significantly mitigates, parallax differences, and that does not rely on processing techniques that exhibit relatively excessive latency. The present invention addresses at least these needs.